ISSN 2282-5452

WWW.DAGLIANO.UNIMI.IT

CENTRO STUDI LUCA D’AGLIANO

DEVELOPMENT STUDIES WORKING PAPERS

N. 460

February 2020

The Price of Haircuts:

Private and Official Default

Silvia Marchesi*

Tania Masi**

* University of Milano Bicocca and Centro Studi Luca d’Agliano

** University of Milano Bicocca

The Price of Haircuts: Private and O¢ cial Default

Silvia MarchesiUniversity of Milano Bicocca, CefES and Centro Studi Luca D�Agliano

Tania MasiUniversity of Milano Bicocca and CefES

January 2020

Abstract: This paper studies the relationship between sovereign debt default and sovereign creditrisk by taking into account the depth of a debt restructuring and by distinguishing betweencommercial and o¢ cial debt. We take rating agencies as well as bond yield spreads (EMBIG) asmeasures of a country�s creditworthiness. Our results show that defaults with private creditorsseem to involve some reputational costs up to seven years since the last agreement, while o¢ cialdefaulters may even bene�t from the restructuring episodes. Therefore, we �nd evidence thato¢ cial and private defaults may have di¤erent costs and then induce selective defaults.

Keywords: Sovereign defaults, Credit Rating Agencies, bond yield spreads

JEL Classi�cation: F34, G15, G24, H63

Acknowledgement: We would like to thank Nicola Borri, Sergio de Ferra, Julian Donaubauer,Aitor Erce, Andreas Fuchs, Kai Gehring, Christoph Moser, Tiziano Razzolini, Christoph Trebeschand participants to the 12th BBQ Workshop (Kiel 2019), Development Economics and PolicyConference, DIW (Berlin 2019), Economics of Global Interactions Conference (Bari 2019), Eu-ropean Public Choice Society Annual meeting (Jerusalem 2019) and 60th SIE Annual meeting(Palermo 2019) for useful comments and suggestions.

E-mail addresses: silvia.marchesi@unimib.it; tania.masi@unimib.it

1

1 Introduction

The empirical literature on sovereign defaults has generally found that default costs are di¢ cult

to quantify and short lived.1 Only more recently, thanks to a more precise measurement of a

country�s repayment record, more persistent e¤ects of default have be detected, which are more

in line with the e¤ects of a default according to the theory. Little is still known, however, on the

implications of debt restructurings involving o¢ cial creditors, despite the role historically played

in the resolution of debt crises (IMF 2013) and the fact that o¢ cial debt accounts for a substantial

share of total sovereign debt, especially in developing countries.

In this paper, we document the relationship between sovereign debt default and a country�s credit

risk, by taking as dependent variable both an indirect and a direct measure of borrowing costs, such

as agency ratings and bond yield spreads (from J.P. Morgan�s EMBI Global -EMBIG). Compared

to bond spreads, credit ratings are available for a larger set of countries and remain a reliable

measure in times of crisis.

We take into account a measure of creditors�loss (or haircut), as in Cruces and Trebesch (2013a),

and we distinguish between o¢ cial and private restructurings. More speci�cally, o¢ cial restruc-

turing stands for agreements reached with o¢ cial creditors (in the Paris Club of o¢ cial creditors,

hereafter Paris Club). In contrast, private restructuring denotes a restructuring deal with private

creditors (foreign banks and bondholders).

We add to previous works by comparing the rating outcome of o¢ cial and private restructurings,

hence primarily contributing to the emerging empirical literature on o¢ cial debts. To the best

of our knowledge, it is the �rst time in this literature that the distinction between private and

o¢ cial deals, as well as the occurrence and magnitude of a default, are taken into account in the

1This literature has mainly looked at the e¤ects of sovereign defaults on international trade (e.g., Rose 2005,Borensztein and Panizza 2010), international credit market (e.g., Borensztein and Panizza 2009, Gelos et al. 2011and Panizza et al. 2009), and GDP growth (Sturzenegger 2004, Borensztein and Panizza 2009, De Paoli et al.2009, Levy Yeyati and Panizza 2011), �nding, overall, short lived e¤ect of sovereign defaults. For a survey of thisliterature see Panizza et al. (2009) and Tomz and Wright (2013).

2

context of credit risk.

Sovereign credit ratings can be interpreted as a forward-looking summary indicator of macro-

economic and (often) political conditions as these a¤ect repayment prospects and tend to be

highly correlated with borrowing costs.2 We should clarify, however, that these measures explic-

itly pertain to a sovereign�s ability (and willingness) to service �nancial obligations to nono¢ cial

(commercial) creditors. Hence, they are �biased�in favour of measuring the probability of default

on debt owed to private creditors.

Understanding how rating agencies and institutional investors evaluate the repayment ability to-

wards o¢ cial creditors is not straightforward. This depends on how �visible�o¢ cial debt risk is

and on how rating agencies incorporate it into their rating models. From their o¢ cial documenta-

tion, rating agencies seem to evaluate o¢ cial risk only to the extent to which it can also a¤ect the

repayment prospects of government obligations to the private sector, due to the preferred creditor

status associated with many of o¢ cial claims (e.g., DBRS 2018).3

In other words, o¢ cial debt seems to be generally perceived as �problematic�, and hence adversely

a¤ect sovereign rating, only to the extent to which arrears to o¢ cial creditors may indicate growing

�nancial distress and/or lack of willingness to pay, which eventually is going to a¤ect private

repayments as well. What is more, o¢ cial creditors (notably the Paris Club) may directly seek

comparable treatment for private-creditor claims as part of any restructuring of their own claims

(e.g., Fitch 2019).4

2Cantor and Packer (1996) were among the �rst to focus on the relationship between default history and creditratings, �nding that countries that defaulted after 1970 are associated with a signi�cant drop in a country�s creditrating.

3Such preferred status, however, is not con�rmed by a recent paper of Schlegl, Trebesch and Wright (2019),who, while con�rming that multilateral institutions are senior creditors, show that o¢ cial bilateral debt is junior,or at least not senior, to bank loans and bonds.

4The Paris Club�s �comparability of treatment" principle dictates that private creditors (mainly banks, bond-holders and suppliers) should receive �a treatment on comparable terms" to those granted by the Paris Club.Hence, creditor governments expect private creditors to share the debt burden by accepting haircuts that are atleast as high as those negotiated by the Paris Club. Timing is also very important as rating agencies may consideran agreement with the Paris Club a positive (or negative) event depending on whether it is (or not) followed by aprivate deal. In a similar vein, they may positively evaluate a private agreement which is directly followed by ano¢ cial one that may contribute to reducing the overall debt burden.

3

Debt restructuring could a¤ect a country�s prospects in at least two alternative ways. Default

involving higher haircuts/restructurings may entail more severe reputational costs. On the other

hand, the channel of debt relief operates in the opposite direction. Since higher haircuts reduce

the level of government�s debt more substantially, such debt reduction may allow countries to exit

a debt overhang improving in this way economic prospects, as described by Krugman (1988). The

overall impact of a debt restructuring on a country�s economy is then theoretically ambiguous and

remains an empirical question. Showing the heterogeneous determinants of default, as well as the

heterogeneous treatment of creditors in the event of default, is important as it could help to shed

light on what precisely are the costs of default to a sovereign country.5

Analyzing 264 default episodes in 130 countries over the period 1990-2013, and using dyadic

monthly data for agency rating and monthly data for institutional investors�ratings, we �nd that

commercial and o¢ cial defaults are associated with di¤erent outcomes.6 Private defaults seem to

involve some reputational costs up to seven years since the last agreement, while o¢ cial defaulters

may even bene�t from the present value reduction. Using the EMBIG spread as dependent

variable, we con�rm the results of Cruces and Trebesch (2013) in the case of private haircuts,

while we �nd that spreads go down up to seven years after �nal o¢ cial deals.

Even if our results may depend on how rating agencies incorporate o¢ cial risk into their rating

models, they are important because they document that the costs of default vary with the amounts

of debt and the type of creditors a¤ected.7 Consistently with Schlegl et al. (2019), we �nd that

defaulting on private debt is highly visible and then less likely to result in a rating downgrade.

5A recent paper (Arellano, Mateos-Planas and Ríos-Rull 2019) presents a theory of sovereign default able torationalize the large heterogeneity in debt crisis, which are typically partial and vary in their duration.

6Our data allows us to take into account of the dyadic relationship between agency-country pairs, at least astime-invariant factors are concerned. Recent studies document the existence of incentives of ratings agencies todistort ratings in favor of their respective home countries, as well as economically and culturally aligned countries(Fuchs and Gehring 2016) or of issuers, in the market for commercial mortgage-backed securities (Sean Chu andRysman 2019).

7The importance of the way in which restructuring are actually arranged, at least for private defaulters, is alsocon�rmed by the results of both Asonuma and Trebesch (2016), Trebesch and Zabel (2017) and Asonuma et al.(2019), who �nd that less confrontational (or preemptive) restructurings are associated with a lower output loss ascompared to soft (non-preemptive) defaults.

4

On the other hand, an o¢ cial default, which often occurs without much media coverage, is much

less visible and hence less likely to determine some collateral damage.

In particular, o¢ cial restructuring that are arranged within the "Paris club umbrella" are supposed

to guarantee a relatively smoother approach to the way in which deals are actually orchestrated

than private ones, hence lowering even further the collateral damage of a default.8 Sovereigns are

aware that the consequences of a default depend in important ways on who the defaulted creditors

are and what bargaining power each creditor group has, hence they may decide to prioritize their

repayments accordingly (e.g., Erce and Mallucci 2018). Documenting this di¤erence can then help

shedding light on why countries default towards whom and why.

After the Greek debt restructuring of 2012, private sovereign debt has been replaced by o¢ cial

debt. More generally, in response to the sovereign debt crises that shook the euro area since the end

of 2009, the governments of Cyprus, Greece, Ireland and Portugal all received o¢ cial funds from

both the International Monetary Fund and newly created European Financial Stability Facility

at �rst, then European Stability Mechanism (ESM). Our results may then provide some insight

also for the debate on the consequences of debt heteorogeneity in Europe, which introduces the

possibility for governments to operate selective defaults discriminating across investors.

This paper contributes to the (empirical) literature of default costs. In particular, to the emerging

literature focusing on the characteristics and the economic relevance of debt restructuring, both

from both a private sector perspective (Asonuma and Trebesch 2016; Forni, Palomba, Pereira

and Richmond 2016; Reinhart and Trebesch 2016; Schlegl et al 2019; Trebesch and Zabel 2017;

Asonuma et al. 2019) and an o¢ cial sector perspective (Cheng, Díaz-Cassou, Erce 2017, 2018a,

2018b; Corsetti and Erce 2018; Marchesi and Masi 2018; Reinhart and Trebesch 2016).9

8As argued by Tomz (2007) concerns about reputation sustain international lending and repayments. Hence, anymeasure that would help to reinforce the reputational mechanism between debtors and creditors are particularlyimportant as they would be to help investors distinguish excusable defaults and inexcusable ones (e.g., Grossmanand Van Huyck 1988).

9Contrarily to Marchesi and Masi (2018), however, here we �nd that higher private haircuts are followed bylower ratings in the post-default period, while Marchesi and Masi (2018) document that higher private haircutsare not associated with lower growth in the aftermath of the debt crisis. As argued by Trebesch and Zabel (2017),

5

The rest of this paper is organized as follows. Section 2 describes the data. Section 3 presents the

empirical model and the results in the dyadic setting of rating agencies, while the results obtained

using the EMBIG bond spread as the dependent variable are presented in Section 4. Section 5

concludes.

2 Data

As a proxy for the severity of the debt restructuring, we consider the corresponding present value

reduction, or "haircut", and (as robustness) the face value reduction.10 We focus on restructurings

with foreign creditors, thus excluding debt restructurings that mainly a¤ected domestic creditors.

Figure 1 shows the evolution over time of the relative shares of o¢ cial and private external debt

for all countries in our sample. As we can see, o¢ cial debt accounts for a substantial share of

total sovereign (external) debt. Moreover, the shares of o¢ cial and private debt have remained

stable over the last forty years. In light of this observation, there is still too little research on the

relative treatment of o¢ cial versus private defaults.

FIGURE 1 HERE

We relied on the original dataset by Cruces and Trebesch (2013b) for the data on debt restruc-

turings with commercial creditors.11 This dataset provides a list of 187 distressed sovereign debt

restructurings with external banks and bondholders that occurred between 1970 and 2013. It

includes information on the amount of debt restructured, the face value reduction, and a measure

of debt relief (Preferred Haircut HSZ ) computed by the authors considering the present value of

both old and new debt instruments.

emerging markets often see a quick recovery of output after �nancial crises, even if credit and capital �ows remaindepressed and market access is unfavourable.10The two measures of private and o¢ cial haircut come from two di¤erent sources and are computed in two

di¤erent ways. For this reason, as a robustness check, we will also consider the private and o¢ cial nominal haircut,which are computed, in both cases, as the ratio of face value debt reduction to the amount of debt treated in therestructuring deal (see Reinhert and Trebesch 2016; and Cheng et al. 2018).11In August 2014, the authors provided an update of their data covering the year 2013 as well.

6

For o¢ cial debt restructurings, we relied on the original dataset built by Cheng et al. (2017),

which contains 429 sovereign debt restructurings with the Paris Club, between 1956 and 2015.12

Paris Club creditors may provide (o¢ cial) debt treatments to debtor countries in the form of

rescheduling (i.e., debt relief by postponement of debt service payments) or, in the case of conces-

sional rescheduling, reduction in debt service obligations during a de�ned period (�ow treatment)

or as of a set date (stock treatment). Following Cheng et al. (2017), by looking at the terms of

treatment, we were able to compute the present value reduction for o¢ cial deals and to compare

this value with the corresponding haircut measure in the case of private agreements (or Preferred

Haircut HSZ) used by Cruces and Trebesch (2013a).13

Our sample includes a maximum of 130 developing countries. Since the data on private debt

restructurings are available only up to 2013, our year sample ends in that year too. It includes 68

defaulting countries, which experienced at least one debt crisis during the sample period as well

as 62 non-defaulters. Among defaulters, 47 countries had both private haircuts and o¢ cial debt

restructurings, 14 countries had only an o¢ cial restructurings (through the Paris Club) while only

7 countries have experienced private defaults only. Table A1 in the Appendix shows all countries

and years, including a list of debt crisis episodes studied here.

12The Paris Club is an informal forum of the most important o¢ cial creditor countries and was designed to dealwith the payment di¢ culties of debtors. The restructuring approach of the Paris Club has evolved over time. Inthe 1980s, negotiations took place on a case-by-case basis and focused on short-term liquidity problems, mostlyimplementing maturity extensions without nominal debt reduction. During the 1990s and 2000s, especially for lowincome countries, restructurings became increasingly concessional, including debt stock cancellations. Speci�cally,as low-income countries are concerned, the possibility of a partial debt stock cancellation of non-ODA debt wasgradually extended from 33% of the eligible debt in 1988 (Toronto Terms) to 50% in 1991 (London Terms) and66% in 1994 (Naples Terms). In 1996, the World Bank and the IMF have implemented the Heavily Indebted PoorCountries (or HIPC) Debt Initiative, which was �rst strengthened in 1999, and, more recently, in 2005, when, underthe Multilateral Debt Relief Initiative (MDRI) multilateral institutions were encouraged to increase their speci�ccontribution to debt reduction. Debt relief at completion point under the HIPC Initiative is provided within theHIPC Exit Terms.13Cheng et al. (2017) provide a detailed overview of the di¤erent terms and report the net present value re-

lief associated with the di¤erent Paris Club Terms of Treatment over the years. We calculated the net presentvalue relief associated with the "ad hoc" agreements by directly looking at the Paris Club documentation.(http://www.clubdeparis.org/en/traitements).

7

Table 1 shows summary statistics for di¤erent subperiods in the full sample of 264 restructurings.14

While the average haircut is about 34 percent over the full sample mean, looking at the three

di¤erent subperiods, we detect a sizeable increase in this amount over time. Average haircut

size is more than double during the last subperiod (2002-2013), as compared to the initial period

(1970-1988), and about 20 percent higher with respect to the intermediate one (1989-2001).

When comparing the size of (private) face value reduction, we can see that there was some nominal

debt reduction in the �rst subperiod in only two cases.15 One reason is that almost all the

settlements up to the beginning of the Brady plan (1989-1994) mainly implied maturity extensions

without face value reduction. Nevertheless, this amount (about 58 percent) exceed, on average,

the reductions granted in the other two subperiods.

As o¢ cial restructurings are concerned, we �nd that the average haircut, over the full period, is

about 64 percent, much lower than the corresponding average private amount.16 Looking at the

three di¤erent subperiods, we also �nd an increase in their size over time. Average haircut size

during the last subperiod (2002-2013) is more than two times the average haircut implemented

during the initial period (1970-1988), and almost double with respect to the average size of the

intermediate period (1989-2001). On the other hand, there is no nominal (o¢ cial) face value

reduction in the �rst subperiod (1979-1988), while the size of o¢ cial nominal relief peaks in the

intermediate years (1989-2001) and slightly decreases in the last period (2002-2013).

As documented by the di¤erent debt relief initiatives, we detect a sizeable and stable amount of

o¢ cial face value reduction over time. Figure 2 shows the evolution over time of the percentage

of both private and o¢ cial debt haircut and face value reduction. As can be seen, while private

agreements were more common up to the mid-nineties, Paris club deals prevail in more recent

14Among those 158 episodes involved restructuring with private creditors, while 106 involved deals with o¢ cialcreditors.15These two episodes refer to the Bolivian buyback and the Mexican "Morgan Bond plan", both taking place in

1988.16As said, in order to compare the two types of defaulters, we only consider o¢ cial restructurings that were

agreed until 2013, which is the last year for which we have information about the size of private restructurings.

8

years. What is more, both haircuts and nominal debt reductions are much higher under o¢ cial

deals.

Table 2 presents summary statistics of the haircut, according to a country�s income. As the

number of countries is concerned, we �nd that middle-income countries tend to default more

with both types of creditors, while low-income countries tend to bene�t from the highest average

haircuts.17 Finally, Figure 3 illustrates the frequency, by size, of haircut and face value debt

reduction. While lower haircuts (and face value reduction) are more common in the case of

private agreements, the opposite holds in the case of o¢ cial restructurings.

TABLES 1 & 2 HERE

FIGURES 2 & 3 HERE

3 Credit Agency ratings

This Section assesses the link between debt crisis and subsequent agency ratings, while in Section 4

the monthly average secondary market bond stripped yield spread (EMBIG) will be the dependent

variable.18

Our main proxy to measure the creditworthiness of a country is its sovereign�s long-term foreign-

currency rating. As shown by Reinhart (2002) ratings predict defaults. Hence this makes them an

informative measure of creditworthiness for countries with severe payment problems. Moreover,

ratings may also represent a ceiling for the credit rating of private companies from the respective

country (Borensztein, Cowan, and Valenzuela 2013). Thus, they may also capture the private

sector�s ease of access to foreign capital (Gehring and Lang 2018). Therefore, they represent a

good proxy for a country�s access to international �nancial markets.19

17The only high-income country which receives an o¢ cial haircut of 100% is Seychelles in 2009.18Data on bond spread, however, are available only for a reduced sample of 47 countries and for the period 1993-

2013. Table A5c, in the Appendix, shows the correlation between the (average) agency rating and bond spread inthis reduced sample.19Afonso, Furceri, and Gomes (2012) related ratings to changes in government bond spreads.

9

We retrieve monthly information, via Bloomberg, on eight rating agencies: CI, Dagong, DBRS,

Fitch, JCR, Moody�s, R&I, and S&P. To analyze the dynamics around default times, we use data

at a monthly frequency. We obtained an unbalanced panel, as each agency assigns ratings to a

di¤erent set of countries over varying time periods. The pair-wise correlation between sovereign

ratings from the eight credit rating agencies under analysis ranges from 0.869 (between Standard

and Poor�s and Dagong Global) and 0.992 (between Fitch and Japan Credit Rating Agency) (see

Table A5b in the Appendix).

For our empirical analysis, all ratings have been translated to a 21-point scale. This means that

we assign the highest value of 21 for an �AAA�rating. �C�and �D�in turn are translated into

a value of one (see Fuchs and Gehring 2017 for a similar approach).

3.1 Method

Since the data on credit agencies are available for the full sample of countries only since 1990,

our monthly data are organized in an unbalanced panel, including a maximum of 130 developing

countries, over the years 1990-2013 (instead of the full period 1970-2013). In order to account for

the possible in�uence of agency-country time-invariant characteristics (what is called the "home

bias" in sovereign rating, see Fuchs and Gehring 2017) we estimate the model using �xed e¤ects

OLS at the agency-country-period-level.20

We estimate the model with agency-country (pair) �xed e¤ects (and cluster the standard errors

at the pair-level), include period-�xed e¤ects, and lag the explanatory variables by one period.

We therefore control for unobserved e¤ects that exclusively vary at the pair and period-level,

substantially reducing concerns over endogeneity. Ordinary least squares treat the dependent

variable as cardinal. This implies that the di¤erence between an �AA�and an �AA+�rating, for

20Fuchs and Gehring (2016) investigates how the home country of rating agencies could a¤ect rating decisions asa result of political economy in�uences and cultural distance. They �nd that agencies have biases in favor of therespective home countries, as well as in favor of culturally more similar countries, and countries in which home-country banks have a larger risk exposure. In particular, cultural proximity (as measured by linguistic similarity)is shown to be the main transmission channel that explains the advantage of the home country.

10

example, is the same as between �BB�and �BB+.�21

In order to identify post-crisis episodes, we focus on ��nal�restructurings only, which we de�ne as

those that were not followed by another restructuring vis-a-vis private or o¢ cial creditors within

the subsequent four years. Moreover, due to our focus on post-restructuring e¤ects, we exclude

observations during default. The information on the duration of private debt crisis come from

Asonuma and Trebesch (for private), while we rely on Beers and Mavalwalla (2018) and Cheng

et al. (2018) for information regarding the duration of o¢ cial debt crises. Following Cruces and

Trebesch (2013a), we take up to seven years after the last haircut, in order to capture the existence

of persistent e¤ects. The regression equation then is:

ci;j;t = �+ �Zi;t�1 + jCi;t�j + �jRi;t�j + �i;j + � t + ui;j;t; j = 1; :::� 3;�4&5;�6&7 (1)

where ci;j;t represents the credit rating provided to country i; by agency j; at period t; Cit�j is

a dummy equal to one when a country has �nalized its �nal private/o¢ cial restructuring and

Rit�j denotes the corresponding amount of private/o¢ cial haircut, and Z is a vector containing

the control variables lagged one period. �i;j and � t denote agency-country pair and year �xed

e¤ects, which allow us to control for both countries time-invariant variation and common trends.

In this way we can also account for global factors that might have in�uenced the simultaneous

dating choice of debt restructuring events (e.g., Baker or Brady plan in the two periods, 1985-88,

or 1989-94). Finally, ui;j;t is the error term.

The advantage of including both o¢ cial and private restructurings in the same speci�cation is

that it allows us to detect their e¤ects by avoiding an omitted variable bias. Moreover, we are also

able to distinguish the rating variation associated with the default per se from that associated

with the amount of the debt a¤ected, i.e. "occurrence" versus "magnitude."

21We should emphasize, however, that the economic consequences of the rating contraction may not be linear,as loosing the two notches from junk territory is clearly di¤erent than switching, for example, from AAA to AA(in S&P�s rating).

11

As the control variables are concerned, we rely on the speci�cation by Cruces and Trebesch (2013a).

Therefore, in order to capture the sovereign�s domestic economic performance, we included public

debt to GDP, the general government net lending/borrowing, GDP real growth, reserves to im-

ports, in�ation rate (based on consumer prices), current account, the ICRG and the political risk

indicator.22 Following Fuch and Gehring (2017), all time-varying control variables enter as lagged

moving averages over one or three years.

Table A2 and Table A5a in the Appendix provides a detailed description of all our variables, while

Table A4a presents some summary statistics.

3.2 Results

Table 3 presents the results obtained by considering the size of private and o¢ cial haircut.23 In

columns 1-2 of Table 3, we include the haircut, expressed in percentage points, up to seven years

after the �nal restructuring (with and without control variables, respectively). In the case of

private deals, the results of Table 3, column 2, show that a one percentage point increase in a

private haircut is associated with a decrease of about 0.05 notch in the credit rating, in year

one after the �nal haircut. This means that a haircut of about 48 percent, which corresponds

to the mean for our sample, can be associated with a decrease of about 2.4 notches in year

one. Accordingly, a one standard deviation increase in haircut (about 25 percentage points in

this sample) is associated with a rating which is 1.25 notch smaller, one year after the private

agreement.

TABLE 3 HERE22As a robustness check, in Table 4, we report the results obtained including further control variables, such as

per capita GDP, total population (in log), and the number of years the chief executive has been in o¢ ce.23The results obtained using the private and o¢ cial face value reduction are reported in the Appendix, in Table

A3, and in Figures A1a and A1b. Due to data limitations, however, we were able to use this variable only in thespeci�cation with agency rating as the dependent variable.

12

Conversely, in the case of an o¢ cial agreement, a one percentage point increase in an o¢ cial

haircut is associated with an increase of about 0.02 notch in the credit rating, in year one after

the restructuring. This means that a haircut of about 45 percent (the mean for our sample) can be

associated with an increase of about 1 notch, in year one. Accordingly, a one standard deviation

increase in an o¢ cial haircut (about 36 percentage points in this sample) is associated with a

rating which is almost one notch higher one year after the o¢ cial deal. When considering the

present value reduction, these results are economically relevant both in the case of private and

o¢ cial deals.

In columns 3-4, we consider only the dummy indicating the occurrence of the private and o¢ cial

haircut, with and without control variables, respectively. Finally, the last two columns contain

the full speci�cation, which includes both the lagged haircut sizes and dummies, with and without

control variables. While all these results are reported for comparison, we mostly base the discussion

on the fully speci�ed model of column 6. It should be kept in mind, however, that the coe¢ cients

shown in the fully speci�ed model have to be interpreted conditionally, as in any interaction model.

The best way to interpret the �ndings of Table 3 is to look at Figure 4a and 4b, which show the

expected variation in agency rating of a restructuring conditional on its size, that is �jRit�j + j,

from equation 1 above.

FIGURES 4a & 4b HERE

Figures 4a and 4b show the expected e¤ect at di¤erent levels of private and o¢ cial haircut,

respectively. The di¤erent panels correspond to the number of years after the restructuring, and

the dotted lines show 90 percent con�dence bands. The e¤ects are calculated from the complete

speci�cation (column 6). Besides easier interpretation, this joint estimate and the resulting graphs

are important because the high correlation between C and R complicates making inference about

their individual e¤ects, but facilitates inference about their sum (see Cruces and Trebesch 2013a).24

24As pointed out by Cruces and Trebesch (2013a), multicollinearity does not bias least squares estimates, butthe high correlation between C and R will tend to increase the estimated standard errors. The high correlation

13

The bottom line of Figure 4a is that private haircuts are negative and statistically signi�cant

in the long term. This can be seen because the upper con�dence band is always below the zero

horizontal line at least for a haircut size greater than 20 percent, which is less than half the average

size in the sample (48 percent). The reduction in credit rating associated with haircut size is also

economically substantial, especially for years four to seven after a restructuring.

In the case of o¢ cial agreements, as shown in Figure 4b, the rating increase of a restructuring is

statistically signi�cant for levels of haircut at which the lower con�dence band is above the zero

horizontal line. In years one to two after the �nal agreements, haircuts greater than 40 percent

(the mean of this sample being about 45 percent) can be associated with signi�cantly higher

ratings. From year three to seven after the restructuring, the rating increase can be signi�cant

only for much larger haircuts (i.e., greater than 60 percent).

The results also hold when we use, as the dependent variable, an average of all the agencies�ratings,

as well as the two separate averages of only American agencies (i.e., Moody�s, Fitch, Standard &

Poor�s, Dominion Bond Rating Services) as opposed to Asian agencies (Dagong Global, Rating

and Investment Information, Japan Credit Rating Agency). They are also robust to using an

ordered-logit model for the discrete 21-step end-of-month rating. Columns 2-4 of Table 4 report

these robustness checks.

TABLE 4 HERE

In summary, the haircut size seems to involve some reputational costs and the correlation be-

tween private restructuring and agency credit rating is negative for years one to seven after the

restructuring episode).25 These results are consistent with Asonuma et al. (2019), who �nd that

post-default restructurings are associated with a decline in bank credit, an increase in lending

interest rates, and a higher likelihood of triggering a banking crisis (especially in the case of

between C and R (about 0.7 in our sample) lowers the variance of the estimated e¤ect of interest, + �R:25This result is in line with Gennaioli et al. (2014) who show that the spillovers of a default, on domestic and

foreign banks, are larger the higher the haircut.

14

pre-emptive agreements).

The opposite holds in the case of o¢ cial agreements, when agency rating generally improve, and

the more so the larger the haircut.26 This evidence then suggests that while for private defaulters

negative spillovers dominate, for o¢ cial defaulters positive (debt relief) spillovers seem to prevail.

In the next Section we will consider a more direct measure of borrowing costs, such as the bond

spread, as in Cruces and Trebesch (2013a).

4 EMBIG spread

In this Section we consider, as the dependent variable, the monthly average secondary market

bond stripped yield spread from J.P. Morgan�s EMBI Global (EMBIG) for each country.27 EMBIG

spreads have been used to proxy foreign currency borrowing costs of both governments and the

private sector in emerging market economies. Due to data availability, the sample is now restricted

to 47 countries over the year 1993-2013. Among the 47 countries covered by the EMBIG, 23 are

defaulters which restructured their debt, while the other 24 countries are non-defaulters.28 Table

A4b presents some summary statistics. We estimate the following equation:

Ei;t = �+ �Zi;t�1 + jCi;t�j + �jRi;t�j + �i + � t + "i;t; j = 1; :::� 3;�4&5;�6&7 (2)

26Since many cases of o¢ cial haircut concern the countries which are eligible for the Heavily Indebted PoorCountry (HIPC) Initiative, these results are, at least to some extent, similar to Raddaz (2011). He �nds thatthe stock prices of South African companies, with subsidiaries in countries bene�ted by multilateral debt reliefinitiatives, increase signi�cantly above those of other �rms, especially around the launching of these initiatives.Similar �ndings were found by Arslanalp and Henry (2005), for the Brady plan (1989-1994).27The stripped yield spread is the di¤erence between the weighted average yield to maturity of a given country�s

bonds included in the index and the yield of a US Treasury bond of similar maturity.28The 23 defaulters are Algeria, Argentina, Belize, Brazil, Bulgaria, Cote d�Ivoire, Croatia, Dominican Republic,

Ecuador, Iraq, Mexico, Nigeria, Pakistan, Panama, Peru, Philippines, Poland, Russia, Serbia and Montenegro,South Africa, Ukraine, Uruguay, and Venezuela. The 24 non-defaulters are: Chile, China, Colombia, Egypt,El Salvador, Gabon, Georgia, Ghana, Greece, Hungary, Indonesia, Jamaica, Kazakhstan, Lebanon, Lithuania,Malaysia, Morocco, South Korea, Sri Lanka, Thailand, Trinidad and Tobago, Tunisia, Turkey and Vietnam. Thislist includes countries with no external sovereign debt restructuring in the chosen period, as well as countries thatentered the EMBIG more than seven years after their restructuring. For more information, see Cruces and Trebesch(2013a).

15

where Ei;t represents monthly bond spread of a country i; at period t; Cit�j is a dummy equal to

one when a country has �nalized its last private/o¢ cial haircut, Rit�j denotes the corresponding

amount of private/o¢ cial haircut in the last restructuring and Z is a vector containing the control

variables (lagged one period). Finally, �i; and � t denote country and time dummies, respectively.

TABLE 5 HERE

The results are presented in Table 5. As in the previous Section, in columns 1-2 of Table 5, we

include the haircut size, expressed in percentage points, up to seven years after the �nal agreement

(with and without control variables, respectively). Column 2 shows that a one percentage point

increase in haircut is associated with EMBIG spreads that are about 3 bp higher in year 1 after

the restructuring. This means that a restructuring involving about 40 percent, which is roughly

the mean for our sample, can be associated with 120 bp higher in year one. Accordingly, a

1 standard deviation increase in amount of debt a¤ected (about 22 percentage points in this

sample) is associated with a rating which is 66 bp higher one year after the private agreement.

On the other hand, in the case of an o¢ cial agreement, a one percentage point increase in an

o¢ cial haircut is associated with a decrease of about 1.8 bp in the credit rating, in year one after

the restructuring. This means that a restructuring with a haircut of about 54 percent (i.e., the

mean for our sample) can be associated with a reduction of almost 100 bp, in year one after the

last o¢ cial agreement. Accordingly, a one standard deviation increase in an o¢ cial haircut (about

44 percentage points in this sample) is associated with a rating which is 80 bp lower one year after

the o¢ cial deal. When considering the present value reduction, these results are economically

relevant both in the case of private and o¢ cial deals.

In columns 3-4, we include only the dummy indicating the occurrence of the private/o¢ cial default,

while the last two columns contain the full speci�cation. As before, we focus on the results for

the fully speci�ed model of column 6, which con�rms the relationship between private haircut and

subsequent spreads for years four to seven after the restructuring.

16

Figures 5a and 5b show the mean increase in bond spreads associated with a debt restructuring, for

di¤erent levels of H and at di¤erent lag lengths, and 90 percent con�dence bands. The �gures are

based on the most demanding speci�cation (column 6 in Table 5). The main message of Figure 5a

is that restructurings with haircuts above 40 percent (the mean of this sample) can be associated

with signi�cantly higher spreads from four to the seven years after a restructuring.29 For further

illustration, suppose that haircuts increase by 1 standard deviation, this implies spreads that

are 145 bp higher in years 4 and 5 after the restructuring, and 132 bp higher in years 6 and 7.

These results are economically relevant and very similar to those obtained by Cruces and Trebesch

(2013a).

Finally, as o¢ cial restructuring are concerned, Figure 5b shows that haircuts above 54 percent

(the mean of this sample) can be associated with signi�cantly lower spreads only from three to

the seven years after the �nal o¢ cial restructuring.30

FIGURES 5a & 5b HERE

As in Cruces and Trebesch (2013a), we �nd that controlling for both the occurrence and the mag-

nitude of default is crucial to detect a more lasting link between debt default and borrowing costs.

Most importantly, private (o¢ cial) restructurings are generally associated with lower (higher) rat-

ings and higher (lower) spreads up to seven years since the last restructuring. As rating and spread

represent indirect and direct measures for borrowing costs, respectively, our result suggest that

the costs of default may vary with the restructuring terms and the relative treatment of o¢ cial

versus private creditors. Our results, therefore, points to the importance of the way in which debt

restructurings are orchestrated, in line with the distinction between "excusable and inexcusable"

(Grossman and van Huyck 1988) and "hard" and "soft" defaults (Trebesch and Zabel 2017).

29In years one and two after the restructuring, the rating variations are only marginally signi�cant.30Hence, the positive growth prospect observed for o¢ cial defaulters after the end of the default (see for example

Marchesi and Masi (2018) might be due to the absence of a negative stigma in the credit markets.

17

5 Conclusions

This paper studies the relationship between sovereign debt default and a country�s creditworthi-

ness, by taking into account the depth of a debt restructuring and by distinguishing between

commercial and o¢ cial sovereign debt agreements. We analyze 264 default episodes in 130 coun-

tries over the period 1990-2013, and we consider agency ratings and bond spreads as indirect

and direct measure of borrowing costs, respectively. Controlling for both the occurrence and the

magnitude of defaults, we �nd a more lasting relationship between debt default and credit risk.

In the case of sovereign ratings, private defaulters are associated with a negative stigma in the

aftermath of the restructuring, while o¢ cial defaulters are overall not a¤ected (or they may even

bene�t) by the restructuring episodes. These results are con�rmed by taking the EMBIG bond

spread as dependent variable over a subsample of countries.

Hence, the trade-o¤ concerning the e¤ects of sovereign debt restructurings seems to be associated

with opposite outcomes for private and o¢ cial defaulters. For the former, negative (reputational)

spillovers seem to prevail, while for o¢ cial defaulters the positive spillovers of a debt reduction

are more important. Thus, our results point to the importance of considering the heterogeneous

treatment of creditors in the event of default. Debtor countries, being aware that the consequences

of default depend on who the defaulted creditors are, may then decide to prioritize their repayments

accordingly.

The analysis is, of course, limited in several respects. We do not claim to draw causal inferences

from the empirical analysis, given the nature of the data available. We do emphasize that the

direction of causality in the relationship between sovereign defaults and credit risk raises some

questions. Thus a robust association between debt defaults and lower (higher) ratings (spreads)

can only be indicative of a correlation between the two variables. Thus, we prefer to interpret the

coe¢ cients in the models below as conditional correlations rather than causal e¤ects.

The analysis is of course limited in several respects. We do not claim to draw causal inferences from

18

the empirical analysis, given the nature of the data available. We do emphasize that the direction

of causality in the relationship between sovereign defaults and credit risk raises some questions.

Therefore, a robust association between debt defaults and lower (higher) ratings (spreads) can

only be indicative of a (conditional) correlation between the two variables, rather than causal

e¤ects.

In a companion paper (Marchesi and Masi 2019), we use the Institutional Investor�s index as a

proxy for a country�s creditworthiness, and the Synthetic Control Method to provide some causal

evidence on the relationship between default and credit ratings.31 We �nd further evidence for the

heterogeneity of the economic impact of debt restructurings, con�rming that o¢ cial and private

defaults may have di¤erent costs and then induce selective defaults.

References

[1] Afonso A., D. Furceri, and P. Gomes. 2012. Sovereign Credit Ratings and Financial Markets

Linkages: Application to European Data. Journal of International Money and Finance 31(3):

606�38

[2] Arellano C., X. Mateos-Planas and J.V. Ríos-Rull. 2019. Partial default. NBER Working

Paper 26076.

[3] Arslanalp S. and P.B. Henry. 2005. Is debt relief e¢ cient? Journal of �nance 62 (2): 1017-051.

[4] Asonuma T. and C. Trebesch. 2016. Sovereign Debt Restructurings: Pre-emptive or Post-

Default," Journal of the European Economic Association 14: 175-214.

[5] Asonuma T., M. Chamon, A. Erce, and A. Sasahara. 2019. Costs of Sovereign Defaults:

Restructuring Strategies, Bank Distress and the Capital In�ow-Credit Channel. IMFWorking

Paper 69.

31Due to data limitations, we could only apply this method to the rating provided by International Investor.

19

[6] Beers D. and J. Chambers. 2007. Default Study: Sovereign Defaults At 26-Year

Low, To Show Little Change In 2007. Standard and Poor�s, Global Credit Portal

www.standardandpoors.com/ratingsdirect

[7] Beers D. and J.Mavalwalla. 2018. The BoC-BoE sovereign default database revisited: what�s

new in 2018? Sta¤Working Paper No. 739.

[8] Borensztein E. and U. Panizza. 2009. The costs of sovereign default. IMF Sta¤ Papers 56

(4): 683�741.

[9] Borensztein E., and U. Panizza. 2010. Do Sovereign Defaults Hurt Exporters? Open Economic

Review 21(3): 339-412.

[10] Borensztein, E., K. Cowan, and P. Valenzuela. 2013. Sovereign Ceilings �Lite�? The Impact of

Sovereign Ratings on Corporate Ratings. Journal of Banking and Finance 37(11): 4014�24.

[11] Buchheit L. C., A. Gelpern , M. Gulati, U. Panizza, B. Weder di Mauro and J. Zettelmeye.

2013. Revisiting Sovereign Bankruptcy. Committee on International Economic Policy And

Reform. Brookings-CEPR October 2013.

[12] Cantor R. and F. Packer. 1996. Determinants and Impact of Sovereign Credit Ratings. Eco-

nomic Policy Review, Federal Reserve Bank of New York 2(2), 37�53.

[13] Capital Intelligence. Sovereign Rating Methodology. September 28, 2018.

[14] Cheng, G., Diaz-Cassou, J. and A. Erce. 2017. From Debt Collection to Relief Provision: 60

Years of O¢ cial Debt Restructurings through the Paris Club. European Stability Mechanism

Working Papers 20.

[15] Cheng, G., J. Diaz-Cassou, and A. Erce. 2018a. The Macroeconomic E¤ects of O¢ cial Debt

Restructuring: Evidence from the Paris Club. Working Paper 21, European Stability Mech-

anism. Oxford Economic Papers, forthcoming

20

[16] Cheng, G., J. Diaz-Cassou, and A. Erce. 2018b. O¢ cial Debt Restructurings and Develop-

ment. World Development, 111, 181-195.

[17] Corsetti G., A. Erce and T. Uy. 2018. Debt Sustainability and the terms of O¢ cial Support.

CEPR Discussion Paper 13292

[18] Cruces J.J. and C. Trebesch. 2013a. Sovereign Defaults: The Price of restructurings.American

Economic Journal: Macroeconomics 5: 85-117.

[19] Cruces J.J. and C. Trebesch. 2013b. Sovereign Defaults: The Price of restructurings: Dataset.

American Economic Journal: Macroeconomics. http://dx.doi.org/10.1257/mac.5.3.85.

[20] De Paoli B., Hoggarth G. and V. Saporta. 2009. Output costs of sovereign crises: some

empirical estimates. Bank of England Working Paper 362.

[21] Dell�Ariccia, G., I. Schnabel and J.Zettelmeyer. 2006. How Do O¢ cial Bailouts A¤ect the Risk

of Investing in Emerging Markets? Journal of Money, Credit and Banking 38(7): 1689�1714.

[22] Dominion Bond Rating Services (DBRS). Rating Sovereign Governments - Methodology.

October 2018

[23] Erce A., E. Mallucci. 2018. Selective Sovereign Defaults. Board of Governors of the Federal

Reserve System International Finance Discussion Paper 1239

[24] Fitch Ratings. Sovereign Rating Criteria - Master Criteria. May 27, 2019.

[25] Forni L., Palomba G., Pereira, J. and C. Richmond. 2016. Sovereign Debt Restructuring and

Growth. IMF Working Paper 147.

[26] Fuchs A. and K. Gehring. 2017. The Home Bias in Sovereign Ratings. Journal of the European

Economic Association, 6: 1386�1423.

[27] Gelos G., Sahay R. and G. Sandleris. 2011. Sovereign Borrowing by developing countries:

what determines market Access? Journal of International Economics 83 (2): 243�254.

21

[28] Gehring K. and V. F. Lang. 2018. Stigma or Cushion? IMF Programs and Sovereign Cred-

itworthiness. CESifo Working Paper 7339.

[29] Gennaioli, N., A. Martin, and S. Rossi. 2014. Banks, Government Bonds, and Default: What

do the Data Say? IMF Working Paper 120.

[30] Grossman H.I. and J.B. Van Huyck. 1988, Sovereign Debt as a Contingent Claim: Excusable

Default, Repudiation, and Reputation. The American Economic Review 78 (5): 1088-1097.

[31] International Country Risk Guide. 2018. Country Rankings, The PRS Group, Various Years.

[32] International Monetary Fund. 2018. World Economic Outlook Database. Washington, DC.

[33] International Monetary Fund. 2018. International Financial Statistics. Washington, DC.

[34] Krugman P. 1988. Financing vs. forgiving a debt overhang. Journal of Development Eco-

nomics 29(3): 253-268.

[35] Jianzhong G. 2019. Dagong Guiding Principles of Credit Rating. Mimeo.

[36] Levy Yeyati E. and U. Panizza. 2011. The Elusive Costs of Sovereign Defaults. Journal of

Development Economics 94: 95-105.

[37] Marchesi S. and T. Masi. 2018. Life after default: O¢ cial vs. private sovereign debt restruc-

turings. University of Milano Bicocca DEMS Working Paper 370.

[38] Marchesi S. and T. Masi. 2019. Sovereign risk after sovereign restructurings. Private and

o¢ cial default. Università di Milano Bicocca DEMS-CefES Working Paper 423.

[39] Moody�s Investor Service. Sovereign Rating History. 2013. https://www.moodys.com.

[40] Moody�s Investors Service. Sovereign Bond Ratings - Rating Methodology. November 27,

2018

22

[41] Panizza U., Sturzenegger F. and J. Zettelmeyer. 2009. The Economics and Law of Sovereign

Debt and Default. Journal of Economic Literature 47(3): 1-47.

[42] Raddatz C. 2011. Multilateral debt relief through the eyes of �nancial markets. The Review

of Economics and Statistics, 93(4): 1262�1288

[43] Rating and Investment Information (R&I). R&I�s Analytical Approach to Sovereigns. Febru-

ary 20, 2018.

[44] Reinhart, C. M. and C. Trebesch. 2016. Sovereign Debt Relief and its Aftermath. Journal of

the European Economic Association 14(1): 215-251.

[45] Reinhart C. M. 2002. Default, Currency Crises, and Sovereign Credit Ratings. World Bank

Economic Review 16(2): 151�70.

[46] Reinhart, C.M., K. S. Rogo¤, and M. A. Savastano. 2003. Debt Intolerance. Brookings Papers

on Economic Activity: 1: 1-74.

[47] Rose A. K. 2005. One Reason Countries Pay Their Debts: Renegotiation and International

Trade. Journal of Development Economics 77: 189-206.

[48] Schlegl M., C. Trebesch and M.L.J. Wright, 2019, �The Seniority Structure of Sovereign

Debt�NBER Working Paper 25793.

[49] Sean Chu C. and M. Rysman. 2019. Competition and Strategic Incentives in the Market for

Credit Ratings: Empirics of the Financial Crisis of 2007. American Economic Review 109(10):

3514�3555

[50] Standard & Poor�s (S&P). How We Rate Sovereigns. February 15, 2019.

[51] Sturzenegger F. 2004. Tools for the Anlysis of debt problems. Journal of Reconstructing

Finance 1(1): 1-23.

23

[52] Sturzenegger F. and J. Zettelmeyer. 2008. restructurings: Estimating Investor Losses in Sov-

ereign Debt Restructurings, 1998-2005. Journal of International Money and Finance 27: 780-

805.

[53] Tomz M. 2007. Reputation and International Cooperation. Sovereign Debt across Three Cen-

turies. Princeton University Press

[54] Tomz M. and M. L. J. Wright. 2013 Empirical Research On Sovereign Debt And Default.

NBER Working Paper 18855.

[55] Trebesch C. and M. Zabel. 2017. The Output Costs of Hard and Soft Sovereign Default. The

European Economic Review 92: 416-432

[56] World Bank. 2017. Database of Political Institutions. Washington, DC.

[57] World Bank. 2018. World Development Indicators. Washington, DC.

24

Tables and �gure

25

26

Table 1: Restructurings and Haircuts over time (in %) Observations Mean SD Min Max

Private haircut 1970-1988 81 23 53 -10 93 1989-2001 57 43 26 -8 92 2002-2013 20 53 31 5 96 Official haircut 1970-1988 1 33 0 33 33 1989-2001 71 58 20 12 100 2002-2013 34 77 28 4 100 Private face value reduction 1970-1988 2 58 40 30 86 1989-2001 34 41 30 1 92 2002-2013 14 56 30 4 96 Official face value reduction 1970-1988

1989-2001 13 74 30 14 100 2002-2013 30 63 25 17 100

Table 2: Restructurings and Haircuts by country's income

Private Haircut (Average size %) High Income Middle Income Low Income 27 33 53

Official Haircut (Average size %) High Income Middle Income Low Income 100 65 62

Private Haircut (# of countries) High Income Middle Income Low Income 7 42 5

Official Haircut (# of countries) High Income Middle Income Low Income 1 22 9

Private Face Value Reduction (Average size %) High Income Middle Income Low Income 38 41 91

Official Face Value Reduction (Average size %) High Income Middle Income Low Income 45 56 80

Private Face Value Reduction (# of countries) High Income Middle Income Low Income 4 30 4

Official Face Value Reduction (# of countries) High Income Middle Income Low Income 1 13 9

27

Table 3: Private and Official Haircut and Agency credit rating, 1990‐2013, OLS

(1) (2) (3) (4) (5) (6) Final Private Haircut (-1) -0.064*** -0.046*** -0.030* -0.034**

(-4.901) (-4.505) (-1.679) (-2.491) Final Private Haircut (-2) -0.048*** -0.032*** -0.019 -0.028**

(-4.658) (-3.801) (-1.255) (-2.366) Final Private Haircut (-3) -0.028*** -0.018*** -0.003 -0.011

(-5.107) (-3.707) (-0.344) (-1.363) Final Private Haircut (-4 & 5) -0.023*** -0.017*** -0.007 -0.014**

(-4.610) (-3.677) (-0.775) (-2.001) Final Private Haircut (-6 & 7) -0.015*** -0.013*** -0.012 -0.012**

(-3.404) (-3.532) (-1.468) (-2.151) Final Official Haircut (-1) 0.001 0.021*** 0.032*** 0.025***

(0.165) (5.851) (3.260) (3.880) Final Official Haircut (-2) -0.000 0.015*** 0.030*** 0.021***

(-0.067) (4.617) (3.464) (3.146) Final Official Haircut (-3) -0.003 0.009** 0.022*** 0.017**

(-0.700) (2.585) (2.694) (2.525) Final Official Haircut (-4 & 5) -0.002 0.009*** 0.018*** 0.019***

(-0.571) (3.051) (2.730) (3.723) Final Official Haircut (-6 & 7) 0.001 0.008*** 0.010*** 0.014***

(0.394) (2.777) (2.693) (3.948) Final Priv. Haircut Dummy (-1) -3.117*** -2.458*** -1.879*** -0.828*

(-5.768) (-4.777) (-3.024) (-1.809) Final Priv. Haircut Dummy (-2) -2.239*** -1.444*** -1.522*** -0.243

(-5.593) (-3.584) (-2.901) (-0.506) Final Priv. Haircut Dummy (-3) -1.317*** -0.756*** -1.242*** -0.389

(-5.072) (-2.849) (-2.733) (-0.908) Final Priv. Haircut Dummy (-4 & 5) -0.948*** -0.635*** -0.719* -0.120

(-4.376) (-2.732) (-1.847) (-0.335) Final Priv. Haircut Dummy (-6 & 7) -0.504*** -0.421** -0.124 -0.028

(-2.756) (-2.494) (-0.390) (-0.107) Final Off. Haircut Dummy (-1) -0.647 0.997*** -2.183*** -0.252

(-1.414) (2.709) (-3.456) (-0.501) Final Off. Haircut Dummy (-2) -0.792** 0.513 -2.097*** -0.465

(-2.080) (1.534) (-3.995) (-0.916) Final Off. Haircut Dummy (-3) -0.771** 0.186 -1.713*** -0.609

(-2.440) (0.651) (-3.884) (-1.348) Final Off. Haircut Dummy (-4 & 5) -0.687** 0.073 -1.475*** -0.795**

(-2.515) (0.276) (-4.065) (-2.479) Final Off. Haircut Dummy (-6 & 7) -0.266 0.113 -0.705*** -0.538***

(-1.537) (0.728) (-3.361) (-3.715) GDP real growth (-1) 0.044*** 0.047*** 0.046***

(3.263) (3.333) (3.370) Primary balance to GDP (-1) 0.003 0.002 0.004

(0.206) (0.133) (0.243) Current Account to GDP (-1) -0.030*** -0.030*** -0.029***

(-3.157) (-3.142) (-3.025) Reserves to imports (-1) 0.003 0.003 0.002

(0.995) (1.079) (0.811) Public debt to GDP (-1) -0.045*** -0.045*** -0.044***

(-5.218) (-5.036) (-5.082) Inflation (-1) 0.581 0.586 0.567

(0.357) (0.356) (0.347) (Absence of) Political risk (-1) 0.159*** 0.153*** 0.157***

(8.562) (8.263) (8.268) Constant 13.135*** 5.503*** 13.225*** 5.807*** 13.213*** 5.566***

(18.575) (3.264) (18.614) (3.456) (18.596) (3.292) Observations 57,984 43,616 57,984 43,616 57,984 43,616 R-squared 0.134 0.394 0.142 0.387 0.151 0.396 Number of pair_id 454 363 454 363 454 363 Pair FE YES YES YES YES YES YES Period FE YES YES YES YES YES YES

Notes: This table shows coefficients of an unbalanced panel data regression with OLS fixed effects at the agency-country-period-level. Agency-country and period-fixed effects are included. Standard errors are clustered at the agency-country-level, t statistics are in parentheses. Significance levels: *0.10, ** 0.05, *** 0.01. The dependent variable is the, monthly country agency rating, while the key explanatory variables are the lagged values of C and R taken up to seven years after each final restructuring.

28

Table 4: Private and Official Haircut and Agency rating, 1990‐2013, OLS (1) (2) (3) (4) (5) Final Private Haircut (-1) -0.032** -1.425** -0.020 -0.171*** -3.328***

(-2.431) (-2.247) (-1.409) (-10.979) (-2.790) Final Private Haircut (-2) -0.027** -0.659 -0.018 -0.162*** -1.770

(-2.371) (-0.942) (-1.232) (-6.179) (-1.312) Final Private Haircut (-3) -0.012 -0.880 -0.007 -0.131*** -1.624

(-1.537) (-1.369) (-0.596) (-3.481) (-0.983) Final Private Haircut (-4 & 5) -0.014* -0.284 -0.011 -0.135*** -0.128

(-1.880) (-0.490) (-1.069) (-3.680) (-0.094) Final Private Haircut (-6 & 7) -0.012** 0.017 -0.013 0.034*** 0.399

(-2.129) (0.037) (-1.404) (3.025) (0.369) Final Official Haircut (-1) 0.022*** -0.018 0.020* -0.012

(3.409) (-1.265) (1.750) (-0.667) Final Official Haircut (-2) 0.020*** -0.016 0.016 -0.013

(2.811) (-1.129) (1.327) (-0.627) Final Official Haircut (-3) 0.016** -0.002 0.014 0.006 0.002

(2.314) (-0.168) (1.102) (0.912) (0.080) Final Official Haircut (-4 & 5) 0.019*** -0.012 0.016* 0.005 -0.024

(3.626) (-1.118) (1.742) (1.173) (-0.982) Final Official Haircut (-6 & 7) 0.013*** -0.013 0.014*** 0.001 -0.027

(3.765) (-1.445) (2.673) (0.354) (-1.273) Final Priv. Haircut Dummy (-1) -0.926** 0.097 -1.250* 0.427 -0.411

(-2.008) (0.095) (-1.923) (0.363) (-0.154) Final Priv. Haircut Dummy (-2) -0.271 -0.275 -0.439 1.350 -1.110

(-0.563) (-0.294) (-0.557) (1.185) (-0.443) Final Priv. Haircut Dummy (-3) -0.297 -0.554 -0.568 1.377 -1.298

(-0.691) (-0.762) (-0.861) (1.042) (-0.781) Final Priv. Haircut Dummy (-4 & 5) -0.138 -0.751 -0.207 1.523 -1.451

(-0.373) (-1.386) (-0.347) (1.332) (-1.232) Final Priv. Haircut Dummy (-6 & 7) -0.031 -0.755** 0.028 -0.958* -1.267*

(-0.116) (-2.265) (0.060) (-1.969) (-1.804) Final Off. Haircut Dummy (-1) -0.060 0.018 -0.061 -0.947 0.040

(-0.122) (1.363) (-0.070) (-0.972) (1.168) Final Off. Haircut Dummy (-2) -0.343 0.015 -0.325 -1.390** 0.035

(-0.650) (1.186) (-0.356) (-2.353) (1.059) Final Off. Haircut Dummy (-3) -0.517 0.012 -0.594 -1.306*** 0.023

(-1.112) (1.047) (-0.685) (-2.744) (0.925) Final Off. Haircut Dummy (-4 & 5) -0.805** 0.015 -0.733 -1.185*** 0.028

(-2.453) (1.599) (-1.241) (-3.646) (1.288) Final Off. Haircut Dummy (-6 & 7) -0.475*** 0.014** -0.627*** -0.332** 0.021

(-3.237) (2.182) (-2.681) (-2.069) (1.498) GDP real growth (-1) 0.171*** 0.036 0.043* 0.047 0.027

(3.191) (1.624) (1.885) (1.360) (0.859) Primary balance to GDP (-1) 0.001 0.011 0.019 -0.058 0.030

(0.066) (0.492) (0.810) (-1.110) (0.663) Current Account to GDP (-1) -0.030*** -0.030** -0.029** -0.011 -0.056*

(-3.130) (-2.112) (-2.226) (-0.258) (-1.905) Reserves to imports (-1) 0.003 0.001 0.003 -0.003 -0.002

(0.925) (0.234) (0.548) (-0.300) (-0.232) Public debt to GDP (-1) -0.043*** -0.040*** -0.038*** -0.054* -0.082***

(-5.014) (-3.196) (-2.890) (-1.811) (-3.392) Inflation (-1) 0.527 1.287 1.049 -5.871 -0.435

(0.323) (0.475) (0.406) (-1.272) (-0.070) (Absence of) Political risk (-1) 0.159*** 0.141*** 0.150*** 0.157** 0.231***

(8.298) (5.005) (5.085) (2.325) (4.584) Change in government -0.317*** (-3.907) Population -0.000 (-0.550) Growth -0.127** (-2.484) Constant 5.221*** 5.469** 4.992* 5.909 (3.065) (2.174) (1.923) (1.240) Observations 43,424 12,937 12,903 5,297 12,937 R-squared 0.403 0.411 0.406 0.538 Number of id 359 84 83 58 84

Notes: In column 1 the regressions are estimated using fixed effects OLS at the agency-country-year-level, (s.e. are clustered at the agency-country-level). In columns 2-4, the regressions are estimated using fixed effects OLS at the country-year-level (s.e. clustered at the country-level). In column 5, the regression is estimated using an ordered-logir model (s.e. are clustered at the country level. The dependent variables are: the dyadic monthly rating (column 1); the monthly mean of all agencies’ rating (column 2 and 5); the monthly mean of the four North American Agencies, i.e., Standard & Poor's, Moody's, Fitch, Dominion Bond Rating Services (column 3); the monthly mean of the three Asian Agencies, i.e., Dagong Global, Rating, Investment Information, Japan Credit Rating Agency (column 4). t statistics are in parentheses. Significance levels: *0.10, ** 0.05, *** 0.01).

29

Table 5: Private and Official Haircut and bond spread, 1990‐2013, OLS (1) (2) (3) (4) (5) (6) Final Private Haircut (-1) 4.724* 2.784 6.177 2.269

(1.762) (1.322) (1.589) (0.872) Final Private Haircut (-2) 3.732* 3.000* 6.087 3.777

(1.824) (1.888) (1.664) (1.154) Final Private Haircut (-3) 3.275 1.404 4.878 3.570

(1.663) (1.131) (1.337) (1.299) Final Private Haircut (-4 & 5) 3.291*** 2.826*** 7.145*** 6.626***

(2.749) (3.711) (2.834) (2.796) Final Private Haircut (-6 & 7) 1.416 1.922** 8.160*** 6.015**

(1.146) (2.252) (3.721) (2.630) Final Official Haircut (-1) -3.482*** -1.781* -4.301*** -2.108

(-3.146) (-1.883) (-2.833) (-0.824) Final Official Haircut (-2) -4.423*** -2.441* -6.393*** -3.995

(-2.798) (-1.717) (-3.023) (-1.234) Final Official Haircut (-3) -3.850** -3.012*** -3.157* -1.798

(-2.376) (-3.617) (-1.727) (-0.589) Final Official Haircut (-4 & 5) -4.216*** -3.793*** -3.859** -3.727

(-4.498) (-5.792) (-2.502) (-1.341) Final Official Haircut (-6 & 7) -3.470*** -1.488*** -1.924 -0.697

(-3.002) (-4.016) (-1.264) (-0.402) Final Priv. Haircut Dummy (-1) 153.344 109.244 -124.943 0.349

(1.416) (1.170) (-0.979) (0.004) Final Priv. Haircut Dummy (-2) 108.032 103.066 -151.820 -64.137

(1.388) (1.502) (-1.291) (-0.491) Final Priv. Haircut Dummy (-3) 89.847 14.486 -106.076 -132.969

(1.056) (0.251) (-0.749) (-1.126) Final Priv. Haircut Dummy (-4 & 5) 68.665 51.534 -204.243 -192.159

(1.017) (0.861) (-1.633) (-1.462) Final Priv. Haircut Dummy (-6 & 7) -24.053 14.458 -304.426*** -181.834

(-0.419) (0.284) (-3.202) (-1.616) Final Off. Haircut Dummy (-1) -83.458 -25.141 159.246 111.896

(-0.384) (-0.187) (1.628) (0.629) Final Off. Haircut Dummy (-2) 14.007 -9.020 309.909 191.847

(0.060) (-0.059) (1.666) (0.870) Final Off. Haircut Dummy (-3) -98.585 -122.897 74.600 22.056

(-0.516) (-1.098) (0.742) (0.121) Final Off. Haircut Dummy (-4 & 5) -101.383 -86.462 98.741 97.595

(-0.666) (-0.660) (0.873) (0.531) Final Off. Haircut Dummy (-6 & 7) -128.873 -69.479 -100.634 -68.184

(-1.281) (-0.763) (-1.128) (-0.490) GDP real growth (-1) -4.856 -4.474 -4.134

(-1.540) (-1.489) (-1.481) Primary balance to GDP (-1) -16.698*** -17.488*** -16.806***

(-3.266) (-3.538) (-3.224) Current Account to GDP (-1) -10.003*** -9.764*** -9.525***

(-2.963) (-2.737) (-2.751) Reserves to imports (-1) -1.328 -1.561 -1.165

(-0.999) (-1.067) (-0.690) Public debt to GDP (-1) 9.834*** 10.510*** 9.349***

(3.512) (3.347) (3.542) Inflation (-1) -0.093 -0.038 -0.141*

(-0.936) (-0.334) (-1.832) (Absence of) Political risk (-1) -7.709*** -7.249** -6.429**

(-2.764) (-2.257) (-2.380) Constant 363.940*** 951.647*** 364.566*** 956.923*** 434.986*** 942.554***

(5.245) (3.947) (4.121) (3.465) (4.604) (3.593) Observations 5,115 3,935 5,115 3,935 5,115 3,935 R-squared 0.344 0.455 0.330 0.444 0.364 0.465 Number of country_id 46 34 46 34 46 34 Country FE YES YES YES YES YES YES Year FE YES YES YES YES YES YES

Notes: This table shows coefficients of an unbalanced panel data OLS regression with fixed effects at the country-year-level and country-year clustered standard errors. t statistics are in parentheses. Significance levels: *0.10, ** 0.05, *** 0.01. The dependent variable is the monthly average country yield spread over US Treasury bonds (EMBIG stripped spread) measured in basis points (bp).

30

Figure 1: Share of private and official debt over time

Figure 2: Share of private and official haircut and face value reduction over time

0.2

.4.6

.8S

har

e of

pri

vate

and

off

icia

l deb

t

1970 1975 1980 1985 1990 1995 2000 2005 2010

Private debt Official debt

1980 1985 1990 1995 2000 2005 2010

Haircut

Private Haircut Official Haircut

1990 1995 2000 2005 2010

Face Value Reduction

Private FVR Official FVR

31

Figure 3: Frequency by size of private and official restructurings, haircuts and face value reduction

05

1015

Per

cent

0 20 40 60 80 100Haircut size (in %)

Private

05

1015

2025

Per

cent

0 20 40 60 80 100Haircut size (in %)

Official

Haircut

05

1015

2025

Per

cent

0 20 40 60 80 100Face Value Reduction (in %)

Private0

1020

30P

erce

nt

20 40 60 80 100Face Value Reduction (in %)

Official

Face Value Reduction

32

Figure 4a: Expected effect on agency rating for different levels of private haircut

Notes: Each graph shows the marginal effect of private haircut on agency rating, for different haircut sizes and at different lag lengths. The dashed lines show 90 percent confidence bands. The effects are calculated using the coefficients from Table 3, column 6. The rating contraction of a restructuring is statistically significant for levels of haircut at which the upper confidence band is below the zero horizontal line. We can see that haircut greater than 20 percent (the mean of this sample being about 48 percent) can be associated with significantly lower ratings during the seven years after a restructuring.

Figure 4b: Expected effect on agency rating for different levels of official haircut

Notes: Each graph shows the marginal effect of official haircut on agency rating, for different haircut sizes and at different lag lengths. The dashed lines show 90 percent confidence bands. The effects are calculated using the coefficients from Table 3, column 6. The rating increase of a restructuring is statistically significant for levels of haircut at which the lower confidence band is above the zero horizontal line. From year one to two years after the agreements, we can see that haircut greater than 40 percent (the mean of this sample being about 45 percent) can be associated with significantly higher ratings. From year three to seven years after the restructuring, the rating increase can be significant only for much larger haircuts, i.e., greater than 60 percent.

-6-4

-20

Rat

ing

varia

tion

depe

ndin

g on

Hai

rcu

t

0 20 40 60 80 100Private Haircut (in %)

One year after final Haircut

-4-3

-2-1

01

Rat

ing

varia

tion

depe

ndin

g on

Hai

rcu

t

0 20 40 60 80 100Private Haircut (in %)

2 years after final Haircut

-2-1

.5-1

-.5

0.5

Rat

ing

varia

tion

dep

end

ing

on H

airc

ut

0 20 40 60 80 100Private Haircut (in %)

3 years after final Haircut

-2-1

.5-1

-.5

0.5

Rat

ing

varia

tion

depe

ndin

g on

Hai

rcu

t

0 20 40 60 80 100Private Haircut (in %)

4&5 years after final Haircut

-2-1

.5-1

-.5

0.5

Ra

ting

varia

tion

depe

ndin

g on

Hai

rcu

t

0 20 40 60 80 100Private Haircut (in %)

6&7 years after final Haircut

-10

12

3R

atin

g va

riatio

n de

pend

ing

on H

airc

ut

0 20 40 60 80 100Official Haircut (in %)

One year after final Haircut

-10

12

Rat

ing

varia

tion

depe

ndin

g on

Hai

rcu

t

0 20 40 60 80 100Official Haircut (in %)

2 years after final Haircut

-2-1

01

2R

atin

g va

riatio

n de

pend

ing

on H

airc

ut

0 20 40 60 80 100Official Haircut (in %)

3 years after final Haircut

-2-1

01

2

Rat

ing

varia

tion

depe

ndin

g on

Hai

rcu

t

0 20 40 60 80 100Official Haircut (in %)

4&5 years after final Haircut

-1-.

50

.51

1.5

Rat

ing

varia

tion

depe

ndin

g on

Hai

rcu

t

0 20 40 60 80 100Official Haircut (in %)

6&7 years after final Haircut

33

Figure 5a: Expected effect on bond spread for different levels of private haircut

Notes: Each graph shows the marginal effect of private haircut on agency rating, for different haircut sizes and at different lag lengths. The dashed lines show 90 percent confidence bands. The effects are calculated using the coefficients from Table 7, column 6. The spread increase of a restructuring is statistically significant for levels of haircut at which the lower confidence band is above the zero horizontal line. We can see that haircuts above 40 percent (the mean of this sample) can be associated with significantly higher spreads from four to the seven years after a restructuring.

Figure 5b: Expected effect on bond spread for different levels of official haircut

Notes: Each graph shows the marginal effect of official haircut on agency rating, for different haircut sizes and at different lag lengths. The dashed lines show 90 percent confidence bands. The effects are calculated using the coefficients from Table 7, column 6. The spread decrease of a restructuring is statistically significant for levels of haircut at which the upper confidence band is below the zero horizontal line. We can see that haircuts above 40 percent (the mean of this sample) can be associated with significantly lower spreads from three to the seven years after a restructuring.

-200

020

040

060

080

0

Spr

ead

varia

tion

dep

end

ing

on

Hai

rcut

0 20 40 60 80Private Haircut (in %)

One year after final Haircut

-200

020

040

060

0

Spr

ead

varia

tion

dep

end

ing

on

Hai

rcut

0 20 40 60 80Private Haircut (in %)

2 years after final Haircut

-400

-200

020

040

0

Spr

ead

varia

tion

dep

end

ing

on

Hai

rcut

0 20 40 60 80Private Haircut (in %)

3 years after final Haircut

-400

-200

020

040

060

0

Spr

ead

varia

tion

dep

end

ing

on

Hai

rcut

0 20 40 60 80 100Private Haircut (in %)

4&5 years after final Haircut

-400

-200

020

040

0

Spr

ead

varia

tion

dep

end

ing

on

Hai

rcut

0 20 40 60 80Private Haircut (in %)

6&7 years after final Haircut

-400

-200

020

040

0

Spr

ead

varia

tion

dep

end

ing

on

Hai

rcut

0 20 40 60 80 100Official Haircut (in %)

One year after final Haircut

-500

050

0S

prea

d va

riatio

n d

epen

din

g o

n H

airc

ut

0 20 40 60 80 100Official Haircut (in %)

2 years after final Haircut

-600

-400

-200

020

0

Spr

ead

varia

tion

dep

end

ing

on

Hai

rcut

0 20 40 60 80 100Official Haircut (in %)

3 years after final Haircut

-600

-400

-200

020

040

0

Spr

ead

varia

tion

dep

end

ing

on

Hai

rcut

0 20 40 60 80 100Official Haircut (in %)

4&5 years after final Haircut

-400

-300

-200

-100

010

0S

prea

d va

riatio

n d

epen

din

g o

n H

airc

ut

0 20 40 60 80 100Official Haircut (in %)

6&7 years after final Haircut

Appendix

34

35

Table A1a: Country sample, defaulters Private restructurings Official restructurings Albania 1991-1995 1993-2000

Angola 1989

Argentina 1982-1993 2001-2005 1985-1992 2014 Belize 2006-2013

Benin 1989-2003

Bolivia 1980-1993 1986-2001

Bosnia Herzegovina 1992-1997 1998-2000 Brazil 1983-1994 1983-1992

Bulgaria 1990-1994 1991-1994

Burkina Faso 1991-2002

Cambodia 1995

Cameroon 1985-2003 1989-2006

Chile 1983-1990 1975-1987

Congo, Dem. Rep. 1975-1989 1976-1989 2002-2010 Congo, Rep. 1983-1988 2007 1986-2004 2010 Costa Rica 1981-1990 1983-1993

Cote d'Ivoire 1983-1998 2000-2012 1984-1994 1998-2012 Croatia 1992-1996 1995 Cuba 1983-1985 1985-1986

Dominican Republic 1982-1994 2004-2005 1985-1991 2004-2005 Ecuador 1982-1995 1999-2000 2008-2009 1983-2003

Egypt, Arab Rep. 1987-1991

El Salvador 1990

Ethiopia 1990-1996 1992-2004

Gabon 1986-1994 1987-1995 2000-2004 Georgia 2001-2004

Ghana 1996-2004

Greece 2012

Grenada 2004-2005 2006

Guatemala 1993

Honduras 1981-2001 1990-2005

Indonesia 1994-2005

Iraq 1986-2006 Jamaica 1977-1990 1984-1993

Jordan 1989-1993 1989-2002

Kenya 1992-1998 1994-2004

Kyrgyz Republic 2002-2005

Macedonia 1983-1988 1992-1997 1984-1988 1995-2000 Mali 1988-2003

Mexico 1982-1990 1983-1989

Moldova 2001-2004 2006

Morocco 1983-1990 1983-1992

Mozambique 1983-1991 2007 1984-2001

Nicaragua 1978-1995 2007 1991-2004

Nigeria 1982-1991 1986-1991 2000-2005 Pakistan 1998-1999 1981 1999-2001 Panama 1984-1996 1985-1990

Paraguay 1986-1993

Peru 1978-1997 1978-1996

Philippines 1983-1992 1984-1994

Poland 1981-1994 1981-1991

Romania 1981-1983 1986 1982-1983

Russia 1991-2000 1993-1999

Rwanda 1998-2005

Senegal 1980-1985 1990-1996 1981-2004

36

Serbia Seychelles 2008-2010

Slovenia South Africa 1985-1993

Sri Lanka 2005

Trinidad and Tobago 1988-1989 1989-1990

Turkey 1976-1982 1978-1980

Uganda 1979-1993 1981-2000

Ukraine 1998-2000 2001

Uruguay 1983-1991 2003

Venezuela, RB 1983-1990

Viet Nam 1982-1997 1993

Zambia 1983-1994 1983-2005 Notes: Countries in bold correspond to are those with only private restructurings, while countries in italics are those with only official restructurings.

Table A1b: Country sample, non-defaulters

Andorra Czech Rep. Lesotho Slovak Rep. Armenia Estonia Libya St. Vincent and the Gren. Aruba Faroe Islands Liechtenstein Suriname Azerbaijan Fiji Lithuania Taiwan The Bahamas French Polynesia Macao Tajikistan Bahrain Gibraltar Malaysia Thailand Bangladesh Hong Kong Maldives Tunisia Barbados Hungary Malta Turkmenistan Belarus India Mauritius Turks and Caicos Islands Bermuda Iran Mongolia United Arab Emirates Botswana Isle of Man Montenegro Uzbekistan Cabo Verde Israel Namibia Cayman Islands Kazakhstan Oman China South Korea Papua New Guinea Colombia Kuwait Qatar Curacao Latvia Saudi Arabia Cyprus Lebanon Singapore

37

Table A2: Variable definitions and sources

Variable Definition Source

DEPENDENT VARIABLE

Sovereign Rating Sovereign rating on a 21-point scale, monthly (8 agencies, see Table A2b) Bloomberg EMBIG spreads Monthly average secondary market bond stripped yield spread, (EMBIG) J.P. Morgan VARIABLES OF INTEREST Private Haircut Private debt haircut, in percent Cruces and Trebesch (2013b) Private Haircut Dummy Dummy =1 in case of a private haircut Built by the author Private Face Value Reduction Private debt face value reduction, percent of treated debt Cruces and Trebesch (2013b) Private Face Value Reduction Dummy Dummy =1 in case of a private face value reduction Built by the author Official Haircut Official debt haircut, in percent Cheng, Diaz-Cassou and Erce (2017) Official Haircut Dummy Dummy =1 in case of an official haircut Built by the author Official Face Value Reduction Official debt face value reduction, percent of treated debt Cheng, Diaz-Cassou and Erce (2017) Official Face Value Reduction Dummy Dummy =1 in case of an official face value reduction Built by the author CONTROL VARIABLES External debt to GDP Ratio of external debt to GDP World Development Indicators, World Bank (2018) GDP growth Per capita GDP (constant 2015 US$), Annual rate of change World Development Indicators, World Bank (2018) Reserves to Imports Ratio of external debt to GDP International Financial Statistics, IMF (2018) Inflation Consumer price index (2010 = 100), Annual rate of change World Development Indicators, World Bank (2018) Net lending/borrowing General government net lending/borrowing World Economic Outlook Database, IMF (2018) Current Account Current account to GDP World Development Indicators, World Bank (2018) Political Risk ICRG Political Risk Index International Country Risk Guide, The PRS Group (2018) Per capita GDP Per capita GDP (constant 2005 US$) World Development Indicators, World Bank (2018) Government change Dummy variable with a value of one Database of Political Institutions, World Bank (2017) (log) Popolation Log of total population World Development Indicators, World Bank (2018)

38

Table A3: Private and Official Face Value Reduction and Agency credit rating, 1990-2013, OLS (1) (2) (3) (4) (5) (6) Final Private FVR (-1) -0.076*** -0.059*** -0.048* -0.056**

(-3.897) (-4.663) (-1.814) (-2.140) Final Private FVR (-2) -0.061*** -0.048*** -0.052** -0.066***

(-3.771) (-4.415) (-2.419) (-3.742) Final Private FVR (-3) -0.037*** -0.031*** -0.037** -0.053***

(-3.833) (-4.573) (-2.413) (-4.445) Final Private FVR (-4 & 5) -0.030*** -0.030*** -0.037*** -0.052***

(-3.398) (-4.446) (-2.674) (-4.773) Final Private FVR (-6 & 7) -0.022*** -0.025*** -0.040*** -0.044***

(-2.971) (-4.457) (-3.143) (-4.714) Final Official FVR (-1) 0.008 0.022*** 0.001 -0.028*

(1.049) (3.851) (0.095) (-1.914) Final Official FVR (-2) 0.004 0.015*** -0.008 -0.017

(0.882) (2.877) (-0.707) (-1.094) Final Official FVR (-3) -0.001 0.006 -0.014 -0.014

(-0.372) (1.227) (-1.126) (-0.870) Final Official FVR (-4 & 5) -0.001 0.010*** -0.010 0.027**

(-0.382) (3.163) (-0.824) (2.469) Final Official FVR (-6 & 7) -0.002 0.005* -0.007 0.012

(-0.677) (1.748) (-0.855) (1.287) Final Priv. FVR Dummy (-1) -3.335*** -2.800*** -1.677* -0.324

(-4.428) (-4.185) (-1.871) (-0.236) Final Priv. FVR Dummy (-2) -2.153*** -1.527*** -0.507 0.999

(-3.873) (-2.836) (-0.876) (1.485) Final Priv. FVR Dummy (-3) -1.001*** -0.584* -0.018 1.152***

(-3.294) (-1.726) (-0.038) (2.724) Final Priv. FVR Dummy (-4 & 5) -0.590** -0.557* 0.370 1.024***

(-2.302) (-1.906) (0.976) (3.089) Final Priv. FVR Dummy (-6 & 7) -0.195 -0.320 0.766** 0.839***

(-0.790) (-1.547) (2.042) (3.259) Final Off. FVR Dummy (-1) 0.417 1.706*** 0.529 3.800***

(1.378) (5.029) (0.580) (3.104) Final Off. FVR Dummy (-2) 0.258 1.134*** 0.985 2.413**

(1.093) (5.070) (0.978) (2.187) Final Off. FVR Dummy (-3) -0.155 0.532** 1.028 1.514

(-0.630) (2.370) (0.939) (1.404) Final Off. FVR Dummy (-4 & 5) -0.216 0.593** 0.737 -1.401*

(-0.830) (2.439) (0.620) (-1.824) Final Off. FVR Dummy (-6 & 7) -0.231 0.312 0.377 -0.681

(-1.208) (1.563) (0.524) (-0.998) GDP real growth (-1) 0.041*** 0.045*** 0.040***

(3.165) (3.247) (3.136) Primary balance to GDP (-1) 0.004 -0.000 -0.000

(0.265) (-0.013) (-0.026) Current Account to GDP (-1) -0.029*** -0.029*** -0.033***

(-3.044) (-3.034) (-3.499) Reserves to imports (-1) 0.003 0.003 0.003

(0.953) (1.022) (0.936) Public debt to GDP (-1) -0.048*** -0.046*** -0.049***

(-5.668) (-5.094) (-5.545) Inflation (-1) 0.238 0.487 -0.007

(0.146) (0.294) (-0.004) (Absence of) Political risk (-1) 0.159*** 0.155*** 0.161***

(8.669) (8.314) (8.588) Constant 13.046*** 5.821*** 13.107*** 5.738*** 13.062*** 5.788***

(18.512) (3.517) (18.459) (3.314) (18.366) (3.413) Observations 57,984 43,616 57,984 43,616 57,984 43,616 R-squared 0.124 0.401 0.116 0.385 0.130 0.408 Number of pair_id 454 363 454 363 454 363 Pair FE YES YES YES YES YES YES Period FE YES YES YES YES YES YES

Notes: This table shows coefficients of an unbalanced panel data regression with OLS fixed effects at the agency-country-period-level. Agency-country and period-fixed effects are included. Standard errors are clustered at the agency-country-level, t statistics are in parentheses. Significance levels: *0.10, ** 0.05, *** 0.01.

39

Table A4a: Descriptive Statistics (agency rating) Variable N Mean SD Min Max Agency rating 43616 12.2 3.77 1 21 Final Priv. Haircut Dummy (-1) 43616 0.01 0.09 0 1 Final Priv. Haircut Dummy (-2) 43616 0.01 0.1 0 1 Final Priv. Haircut Dummy (-3) 43616 0.01 0.12 0 1 Final Priv. Haircut Dummy (-4 & 5) 43616 0.03 0.18 0 1 Final Priv. Haircut Dummy (-6 & 7) 43616 0.04 0.2 0 1 Final Private Haircut (-1) 43616 0.39 5.09 0 95.5 Final Private Haircut (-2) 43616 0.45 5.29 0 95.5 Final Private Haircut (-3) 43616 0.52 5.48 0 95.5 Final Private Haircut (-4 & 5) 43616 1.41 8.95 0 95.5 Final Private Haircut (-6 & 7) 43616 1.58 9.13 0 95.5 Final Off. Restr. Dummy (-1) 43616 0 0.07 0 1 Final Off. Restr. Dummy (-2) 43616 0.01 0.08 0 1 Final Off. Restr. Dummy (-3) 43616 0.01 0.08 0 1 Final Off. Restr. Dummy (-4 & 5) 43616 0.01 0.12 0 1 Final Off. Restr. Dummy (-6 & 7) 43616 0.02 0.13 0 1 Final Official Haircut (-1) 43616 0.23 4.23 0 100 Final Official Haircut (-2) 43616 0.25 4.39 0 100 Final Official Haircut (-3) 43616 0.28 4.67 0 100 Final Official Haircut (-4 & 5) 43616 0.61 6.74 0 100 Final Official Haircut (-6 & 7) 43616 0.72 7.38 0 100 GDP real growth (-1) 43616 4.19 3.55 -15.14 33.72 Primary balance to GDP (-1) 43616 -1.81 5.67 -20.35 43.3 Current Account to GDP (-1) 43616 -0.62 8.77 -46.72 45.45 Reserves to imports (-1) 43616 51.53 39.15 1.11 320.27 Public debt to GDP (-1) 43616 45.77 29.28 2.22 183.07 Inflation (-1) 43616 0.5 0.15 0.12 1 (Absence of) Political risk (-1) 43616 68.78 8.66 37.87 89.13

Notes: Descriptive statistics refer to the specification of Table 3, column 6.

Table A4b: Descriptive Statistics (Bond Spread) Variable N Mean SD Min Max Bond spread 4271 358.32 308.91 13.87 3158.22 Final Priv. Haircut Dummy (-1) 4271 0.02 0.13 0 1 Final Priv. Haircut Dummy (-2) 4271 0.02 0.15 0 1 Final Priv. Haircut Dummy (-3) 4271 0.03 0.17 0 1 Final Priv. Haircut Dummy (-4 & 5) 4271 0.07 0.26 0 1 Final Priv. Haircut Dummy (-6 & 7) 4271 0.08 0.27 0 1 Final Private Haircut (-1) 4271 0.79 6.68 0 76.8 Final Private Haircut (-2) 4271 0.82 6.47 0 76.8 Final Private Haircut (-3) 4271 0.88 6.31 0 76.8 Final Private Haircut (-4 & 5) 4271 2.8 11.74 0 89.4 Final Private Haircut (-6 & 7) 4271 2.81 10.89 0 76.8 Final Off. Restr. Dummy (-1) 4271 0.01 0.09 0 1 Final Off. Restr. Dummy (-2) 4271 0.01 0.08 0 1 Final Off. Restr. Dummy (-3) 4271 0.01 0.1 0 1 Final Off. Restr. Dummy (-4 & 5) 4271 0.03 0.16 0 1 Final Off. Restr. Dummy (-6 & 7) 4271 0.02 0.14 0 1 Final Official Haircut (-1) 4271 0.56 7.23 0 100 Final Official Haircut (-2) 4271 0.22 4.1 0 100 Final Official Haircut (-3) 4271 0.26 4.37 0 93.33 Final Official Haircut (-4 & 5) 4271 1.08 9.09 0 93.33 Final Official Haircut (-6 & 7) 4271 1.23 10.13 0 100 GDP real growth (-1) 4271 4.31 3.64 -15.14 18.29 Primary balance to GDP (-1) 4271 -2.23 3.29 -12.75 8.69 Current Account to GDP (-1) 4271 -0.74 5.44 -20.52 21.18 Reserves to imports (-1) 4271 53.87 33.72 5.12 238.24 Public debt to GDP (-1) 4271 46.19 25.63 3.88 183.07 Inflation (-1) 4271 0.48 0.17 0.12 1 (Absence of) Political risk (-1) 4271 67.3 8.45 40.71 87

Notes: Descriptive statistics refer to the specification of Table 5, column 6.

40

Table A5a: List of Agencies

Variable Observations Countries Years Headquarter Source

Standard & Poor's (S&P) 24621 114 1977-2018 United States Bloomberg Moody's Investors Service 22950 117 1986-2018 United States Bloomberg Fitch Ratings 18596 99 1994-2018 United States/France Bloomberg Dominion Bond Rating Services (DBRS) 1609 20 2006-2018 Canada Bloomberg Dagong Global 6079 67 2010-2018 China Bloomberg Rating and Investment Information (R&I) 6189 28 1998-2018 Japan Bloomberg Japan Credit Rating Agency (JCR) 4041 21 1998-2018 Japan Bloomberg Capital Intelligence (CI) 4884 36 2002-2018 Cyprus/Kuwait Bloomberg

Table A5b: Correlations between Agency credit rating, 1990-2018

(1) (2) (3) (4) (5) (6) (7) (8)

Standard & Poor's (S&P) 1 Moody's Investors Service 0.979 1 Fitch Ratings 0.991 0.987 1 Dominion Bond Rating Services (DBRS) 0.977 0.992 0.988 1 Dagong Global 0.869 0.913 0.907 0.919 1 Rating and Investment Information (R&I) 0.934 0.955 0.957 0.954 0.973 1 Japan Credit Rating Agency FN (JCR) 0.942 0.966 0.968 0.972 0.980 0.992 1 Capital Intelligence (Cyprus) 0.974 0.991 0.988 0.989 0.942 0.979 0.986 1

(obs. 245)

Table A5c: Correlation between Agency and EMBIG spread

(1) (2) (3) Agency rating (mean) 1 EMBIG spread -0.563 -0.4956 1

(obs=7,220)

41

Figure A1a: Expected effect on agency rating for different levels of private face value reduction

Notes: Each graph shows the marginal effect of private face value reduction on agency rating, for different face value reduction sizes and at different lag lengths. The dashed lines show 90 percent confidence bands. The effects are calculated using the coefficients from Table A3, column 6. The rating contraction of a restructuring is statistically significant for levels of nominal haircut at which the upper confidence band is below the zero horizontal line. We can see that haircut greater than 20 percent (the mean of this sample being about 50 percent) can be associated with significantly lower ratings during the seven years after a restructuring.

Figure A1b: Expected effect on agency rating for different levels of official face value reduction

Notes: Each graph shows the marginal effect of official face value reduction on agency rating, for different face value reduction sizes and at different lag lengths. The dashed lines show 90 percent confidence bands. The effects are calculated using the coefficients from Table A3, column 6. The rating increase of a restructuring is statistically significant for levels of nominal haircut at which the lower confidence band is above the zero horizontal line. From year one to two years after the agreements, we can see that any haircut can be associated with significantly higher ratings. From year five to seven years after the agreements nominal haircut greater than 60 percent (which corresponds to mean of this sample) can be associated with significantly higher ratings.

-8-6

-4-2

02

Rat

ing

varia

tion

depe

ndin

g on

FV

R

0 20 40 60 80 100Private Face Value Reduction (in %)

One year after Face Value Reduction

-8-6

-4-2

02

Rat

ing

varia

tion

depe

ndin

g on

FV

R

0 20 40 60 80 100Private Face Value Reduction (in %)

2 years after Face Value Reduction

-6-4

-20

2

Rat

ing

varia

tion

depe

ndin

g on

FV

R

0 20 40 60 80 100Private Face Value Reduction (in %)

3 years after Face Value Reduction

-6-4

-20

2

Rat

ing

varia

tion

depe

ndin

g on

FV

R

0 20 40 60 80 100Private Face Value Reduction (in %)

4&5 years after Face Value Redution

-6-4

-20

2R

atin

g va

riatio

n de

pend

ing

on F

VR

0 20 40 60 80 100Private Face Value Reduction (in %)

6&7 years after Face Value Reduction

02

46

Rat

ing

varia

tion

depe

ndin

g on

FV

R

0 20 40 60 80 100Official Face Value Reduction (in %)

One year after Face Value Reduction

-10

12

34

Rat

ing

varia

tion

depe

ndin

g on

FV

R

0 20 40 60 80 100Official Face Value Reduction (in %)

2 years after Face Value Reduction

-10

12

3

Rat

ing

varia

tion

depe

ndin

g on

FV

R

0 20 40 60 80 100Official Face Value Reduction (in %)

3 years after final Face Value Reduction

-3-2

-10

12

Rat

ing

varia

tion

depe

ndin

g on

FV

R

0 20 40 60 80 100Official Face Value Reduction (in %)

4&5 years after Face Value Reduction

-2-1

01

Rat

ing

varia

tion

depe

ndin

g on

FV

R

0 20 40 60 80 100Official Face Value Reduction (in %)

6&7 years after Face Value Reduction